

Fit indices with inconsistent results 

Message/Author 

Eric Klopp posted on Sunday, November 02, 2008  1:20 pm



Hello, I did a CFA using a 7factor model with correlated latent variables. Every one of the observed variables is specified to load on one latent variable and mesurement errors of the observed variables where set to be uncorrelated. Scaling of the latent variable was done setting the regression weight of an observed variable to one. Before testing the entire model with 7 factors every single factor has undergone a CFA with good model fit and was crossvalidated. After that, the factors we put together to the above mentioned 7factor model Now, I have a problem with the results of the 7factor model. Altough my chi^2 is significant (chi^2=767,149 df=474, p<0,001, N=291) the chi^2/df ratio, SRMR and RMSEA showed good to acceptable model fit (chi^2/df=1,618, SRMR=0,0680, RMSEA=0,046, CI(RMSEA)=[0,040;0,052]). But the CFI is rather low (CFI=0,834). How can this result be explained? Many thanks, Eric 


When you put the factors together, there may be a need for crossloadings that impair the fit of the model. I would start with an EFA of all of the items to see if this might be the case. 

Back to top 

